Constructing a proxy for ability bias

In constructing a suitable proxy for the differential abilities of university graduates within fields of study, we were constrained by the fact that only one wave (2002) contained information on the year in which the respondents left the tertiary education system. Based on this information, a variable that calculated the number of years of university study was constructed. From this it was then possible to compute the average number of years studied per academic discipline, since it is well-known that some

degrees (e.g. Medical and Polytechnics) require more time for graduation than others. A minimum and a maximum time bound for each field was also defined as one standard deviation (-) and (+) the mean year, respectively, as shown in Table A5. Finally, three new dummy variables were added to the dataset, defined as follows:

(i) 1 if individuals‟ time of university study is less than the minimum bound by subject, 0 otherwise.

This is believed to proxy for those individuals who dropped out of their studies (as we have now kept those respondents with incomplete studies in the analysis), and corresponds to 6.47% of the sample of graduates.

(ii) 1 if individuals‟ time of university study is between the minimum and maximum threshold, 0 otherwise. This category should act as surrogate for those people who experienced normal university tenure, and occupies the bulk of the sample of graduates (83.39%).

(iii) 1 if individuals‟ time of university study exceeds the maximum bound, 0 otherwise. This group, which comprises 9.86% of the sample, should refer to students who, for various reasons (partly lack of motivation or lower ability), extended their university life.

As can be observed from Table A6, it is found that, compared to those who experienced normal university tenure, only individuals of the first group (i) suffer from statistically significant lower wages (7-8%). The lack of significance of category (iii) may be possibly explained by the fact that many individuals of this type decide to extend their studies due to concurrent employment in the labour market in various jobs, which offers useful work experience. It is also important to notice that the estimates of Table A6 indicate that despite the inclusion in the regression of the above indicators of ability, the estimated returns to education remain largely unaffected.

Table A5 Average Number of Years of Study by Academic Discipline in

Notes: The min bound is calculated by subtracting one s.d. from the average number of years per field, while the max bound is obtained by adding one s.d.

Table A6 Estimates of Returns to Academic Disciplines in Greece with Controls for Ability, 2002, LFS

Agricultural Science -0.288*** -0.290*** -0.218*** -0.229***

(0.048) (0.048) (0.066) (0.065)

Physics & Maths -0.208*** -0.207*** -0.195*** -0.206***

(0.039) (0.039) (0.045) (0.044)

Law -0.149* -0.154* -0.118** -0.129**

(0.080) (0.080) (0.056) (0.055)

Economics & Business -0.188*** -0.186*** -0.228*** -0.231***

(0.038) (0.038) (0.039) (0.039)

Social Sciences -0.194*** -0.195*** -0.147** -0.154**

(0.065) (0.065) (0.074) (0.075)

Humanities -0.327*** -0.327*** -0.214*** -0.222***

(0.042) (0.042) (0.040) (0.040)

Physical Education -0.291*** -0.289*** -0.181*** -0.187***

(0.045) (0.045) (0.059) (0.059)

Agricultural Science -0.335*** -0.330*** -0.281*** -0.286***

(0.066) (0.065) (0.086) (0.085)

Food Technology -0.181* -0.174* -0.355*** -0.367***

(0.096) (0.093) (0.095) (0.095)

Librarianship 0.000 0.000 -0.373*** -0.387***

(0.000) (0.000) (0.138) (0.139)

Medical-related -0.359*** -0.358*** -0.273*** -0.278***

(0.069) (0.070) (0.041) (0.040)

Notes: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1. All returns are measured relative to a “Secondary” degree. The remaining regression output is available from the authors upon request.

Endnotes

1 The University System includes the Universities, the Polytechnics, the Higher Fine Arts Institute and the Hellenic Open University. There are 20 universities in Greece located in various towns. There are also 14 Technological Education Institutes. The main distinction between AEI and TEI universities are that TEI courses are of shorter duration relative to those offered by AEI, are more practically oriented and the entry requirements are in general lower.

2 More recent figures from the OECD (2008) place the percentage close to 18%.

3 Katsanevas (2002) calculates that Greece has by far the largest ratio of doctors or lawyers per head in the EU.

In the year 2000 one lawyer corresponded to every 338 residents, compared to the average EU ratio of 1:850, while the ratio of doctors per residents in the whole country stands at 1/185 (in Athens it is 1/150), compared to 1/350-400 in the EU (Fyntanidoy, 2001).

4These negative prospects have been confirmed by Katsanevas (2002), who studied “the balance of supply and demand of professions.” In this research the conventional fields of Medicine, Law, and Education were classified as having very negative prospects for the future. At the same time the fields of IT, telecommunications and of new technologies, in general, presented very promising opportunities. Both the European Commission (1996) and the OECD (2005) have also argued that Greek universities are merely producing „degree holders‟ who, in the face of a shrinking public sector, have a higher probability of experiencing unemployment/underemployment.

5 This question becomes even more important in the face of recent evidence that the level of pay of Greek graduates is quite responsive to a tentative rise in unemployment at the aggregate level (Livanos, 2008[b]),

6 The spring quarter is used (following Eurostat guidelines) as it is the one from which the annual employment figures are produced in all EU countries.

7 Earnings are calculated as the net monthly wage that the respondents receive from their main employment inclusive of any extraneous payments (such as Christmas and Easter bonus, annual leave remuneration and other irregular bonuses). Specifically, the level of individual income is measured at the midpoint of the respective income band specified by the Greek LFS. The consistency of the wage information has been corroborated with comparison of the LFS data with other major EU datasets that include Greece, such as the EU-SILC and the European Survey on Working Conditions (ESWC). Finally, it is also important to notice that using nominal rather than real wage terms should not affect the relative ranking of the various academic disciplines, as only the constant term would change in the estimation procedure.

8 Buchinsky (2001) shows that there may be considerable differences between the semiparametric estimates that he proposes and those obtained by a traditional parametric probit model in the selection equation.

9 This matches closely with the 0.26 coefficient reported by Machin and McNally (2007) for Greek men.

10 A formal Heckman-type econometric procedure has been employed that confirms that the estimated returns between the public and private sectors are not subject to selectivity bias. However, identification of the model is achieved only on the basis of non-linearities in the functional forms of the equations. For this reason, the results of the selection model are not presented in the paper, though they are available from the authors upon request.

11 The effect of the remaining variables that are included in the wage equations conforms to the familiar patterns that have been reported in the literature (see Table A1), namely upward-sloping age-earnings profiles that are relatively steeper for males; marriage yielding an wage premium over other marital states; full-time workers (particularly women) enjoying higher remuneration relative to part-timers; and wage rates varying substantially among regions, with residents of the capital of Athens and of surrounding areas enjoying higher returns. The cohort dummies also indicate that younger cohorts in Greece are facing an earnings disadvantage relative to older cohorts, which appears to be more pronounced for females.

12 The rate of self-employment is much lower for the female employed population, close to 8-9%, which is why the analysis has focussed on the male sample in the text. Nevertheless, it is also confirmed that the female wage returns do not suffer from selectivity bias (available from the authors upon request).

13 These results are available from the authors upon request.

14 Individuals born prior to 1950 are neglected, given the limited number of observations that result in small cell sizes by field of study.

15 In the past, almost one-third of the Greek student population studied abroad; in 1975 60000 students were registered in Greek institutions and 30000 abroad. Even today nearly 60000 Greek students continue to study abroad (compared to 370000 who are registered in Greek universities i.e. approximately 14% of the total university graduate population).